Although bioclimatic modelling is often used to estimate potential impacts of likely climate changes, little has been done to assess the reliability and variability of projections. Here, using four niche-based models, two methods to derive probability values from models into presence–absence data and five climate change scenarios, I project the future potential habitats of 1350 European plant species for 2050. All 40 different projections of species turnover across Europe suggested high potential species turnover (up to 70%) in response to climate change. However variability in the potential distributional changes of species across climate scenarios was obscured by a strong variability in projections arising from alternative, yet equally justifiable, niche-based models. Therefore, projections of future species distributions and derived community descriptors cannot be reliably discussed unless model uncertainty is quantified explicitly. I propose and test an alternative way to account for modelling variability when deriving estimates of species turnover (with and without dispersal) according to a range of climate change scenarios representing various socio-economic futures.